Biogeochemistry (2005) 76: 161–185 DOI 10.1007/s10533-005-2856-9
Springer 2005
-1
Tracing sources of streamwater sulfate during snowmelt using S and O isotope ratios of sulfate and 35S activity JAMES B. SHANLEY1,*, BERNHARD MAYER2, MYRON J. MITCHELL3, ROBERT L. MICHEL4, SCOTT W. BAILEY5 and CAROL KENDALL4 1
U.S. Geological Survey, P.O. Box 628, Montpelier, VT 05602, USA; 2Department of Geology and Geophysics, University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada T2N 1N4; 3 State University of New York, College of Environmental Science and Forestry, 1 Forestry Drive, Syracuse, NY 13210, USA; 4U.S. Geological Survey, 345 Middlefield Road, MS 434, Menlo Park, CA 94025, USA; 5USDA Forest Service, Northeastern Research Station, Hubbard Brook Experimental Forest, 234 Mirror Lake Road, Campton, NH 03223, USA; *Author for correspondence (e-mail:
[email protected]; Phone: +1-802-828-4466; fax: +1-802-828-4465) Received 28 April 2004; accepted in revised form 28 February 2005
Key words: Isotopic tracers, Oxygen-18, Sulfate, Sulfur-34, Sulfur-35, Vermont, Sleepers River Abstract. The biogeochemical cycling of sulfur (S) was studied during the 2000 snowmelt at Sleepers River Research Watershed in northeastern Vermont, USA using a hydrochemical and multi-isotope approach. The snowpack and 10 streams of varying size and land use were sampled for analysis of anions, dissolved organic carbon (DOC), 35S activity, and d34S and d18O values of sulfate. At one of the streams, d18O values of water also were measured. Apportionment of sulfur derived from atmospheric and mineral sources based on their distinct d34S values was possible for 7 of the 10 streams. Although mineral S generally dominated, atmospheric-derived S contributions exceeded 50% in several of the streams at peak snowmelt and averaged 41% overall. However, most of this atmospheric sulfur was not from the melting snowpack; the direct contribution of atmospheric sulfate to streamwater sulfate was constrained by 35S mass balance to a maximum of 7%. Rather, the main source of atmospheric sulfur in streamwater was atmospheric sulfate deposited months to years earlier that had microbially cycled through the soil organic sulfur pool. This atmospheric/pedospheric sulfate (pedogenic sulfate formed from atmospheric sulfate) source is revealed by d18O values of streamwater sulfate that remained constant and significantly lower than those of atmospheric sulfate throughout the melt period, as well as streamwater 35S ages of hundreds of days. Our results indicate that the response of streamwater sulfate to changes in atmospheric deposition will be mediated by sulfate retention in the soil.
Introduction Trends and patterns of sulfate concentrations in streamwater have received much attention for several decades because sulfate plays a dominant role in the anthropogenic acidification of surface waters (Reuss and Johnson 1986; Likens and Bormann 1995; Driscoll et al. 2001; Likens et al. 2002). In recent years attention in the USA and Europe has focused on how stream sulfate responds to decreases in atmospheric deposition of sulfate (Stoddard et al. 1999; Clow
162 and Mast 1999). Trend detection in surface waters may be confounded by the presence of sulfate produced by weathering of sulfur-containing minerals in the watershed bedrock (Turk et al. 1993; Mayer et al. 1995a; Alewell et al. 1999; Bailey et al. 2004). Rochelle et al. (1987) concluded that sulfate was in approximate balance in many ecosystems in glaciated areas of eastern North America, but more recent studies have found that the amount of sulfate exported in drainage waters exceeds atmospheric inputs, suggesting the importance of an internal sulfur source (Driscoll et al. 1998; Alewell et al. 1999; Mitchell et al. 2001a, b; Likens et al. 2002). Accounting for geologic sulfate sources is essential in analyzing the response of sulfate in streamwater to changes in atmospheric deposition. Sulfur may be retained in a catchment by adsorption, mineral precipitation, uptake, and assimilatory or dissimilatory bacterial sulfate reduction (BSR). These abiotic and biotic processes can result in a substantial residence time of sulfur in soils (Mayer et al. 1995b; Alewell 2001; Novak et al. 2001), causing a lag in streamwater response to changes in atmospheric sulfate inputs. Stable isotope ratios of sulfur (34S/32S) have been used to assess the relative contributions of atmospheric, pedospheric, and lithospheric sulfur to streamflow (e.g. Krouse and Grinenko 1991; Mitchell et al. 1998). The d34S values of mineral sulfur vary widely (Nielsen et al. 1991), sometimes even within a small geographic area (Bailey et al. 2004). In some areas, the d34S values of atmospheric S also vary and this variation has been used to apportion S sources (Alewell et al. 2000; Mast et al. 2001). In the northeastern USA, however, d34S values of sulfate in atmospheric precipitation are often quite consistent, and average near 5& (Mitchell et al. 1998). If d34S values of precipitation and bedrock are uniform within each source but distinct from one another, and isotope fractionation effects during sulfur transformations are minimal, then d34S values of sulfate in streamwater may be used to quantify atmospheric and weathering sources of S. Isotopic approaches provide constraints on the pools, cycling, and residence time of sulfate that may be applied in models that predict ecosystem response to changes in atmospheric input (Gbondo-Tugbawa et al. 2002). Atmospheric sulfur contains a small quantity of the cosmogenic radioactive isotope 35S (half-life 87 days). Detectable 35S in streamwater indicates the presence of S deposited from the atmosphere within the last 1–2 years (Michel and Naftz 1995; Sueker et al. 1999; Michel et al. 2000). If the decrease in specific activity of 35S (mBq per mg SO4) from precipitation to streamflow is due solely to radioactive decay, it may be used to estimate an average age of streamwater sulfate, and thus to assess how rapidly S cycles through a catchment (Michel et al. 2000). However, specific 35S activity may be diluted by 35S-free sulfate generated from mineral weathering, resulting in an older apparent age of atmospheric S unless the non-atmospheric component is subtracted before the calculation. Much of the atmospheric sulfate entering a watershed is retained in the soil and vegetation. Although d34S values and 35S activity help to distinguish atmospheric and geologic S sources and to estimate ecosystem residence time of S, they usually provide only limited information on S cycling in the soil
163 (Mitchell et al. 2001a; Novak et al. 2001). For estimating the contributions of these soil S sources, d18O values of sulfate (Mayer et al. 1995b; Kester et al. 2003) and more recently d17O values of sulfate (Bao et al. 2001; Lee et al. 2001; Johnson et al. 2001) have proved especially useful. Atmospheric sulfate is enriched in 18O because it derives some of its oxygen atoms from atmospheric O2, which has a d18O value of +23& (Kroopnick and Craig 1972). When atmospheric sulfate undergoes assimilatory bacterial sulfate reduction and reoxidation in the soil, the distinctive high d18O value is lost and reset to a lower d18O value (e.g. Mayer et al. 1995b) because some oxygen atoms in the sulfate are replaced by oxygen from soil water with negative d18O values (Krouse and Mayer 2000). These processes typically have only a small effect on the d34S values of sulfate unless dissimilatory bacterial sulfate reduction occurs (Mitchell et al. 1998). The various processes affecting the isotopic composition of sulfate are depicted in Figure 1. At Sleepers River in northern Vermont, USA, Hornbeck et al. (1997) reported that sulfate export in streamwater was greater than two times atmospheric inputs during 1992–1994, suggesting a geologic S source. From a regional survey, Bailey et al. (2004) reported that Sleepers River had an elevated S content in bedrock (7.9 mg S kg1) with a relatively high d34S value (+8.5&, significantly higher than the average d34S value of precipitation, +5.6&). Three base-flow stream samples had sulfate d34S values of +7.3, +9.1, and +10.5&, closely bracketing the value of the bedrock sample. These initial analyses suggested that isotopic source apportionment of sulfate would be possible at Sleepers River, leading to the current study. The objective of this study was to provide new information on catchment S cycling and dynamic variations of S sources during a hydrologic event using a hydrochemical and multi-isotope approach. We chose to study snowmelt because it is a hydrologically active time when 50% of the annual streamflow typically occurs in a 6-week period. Our approach was to analyze the d34S and d18O values of sulfate and cosmogenic 35S activities in the snowpack, groundwaters, and surface waters to evaluate the relative contribution of different S sources and the residence time of atmospherically derived S in the catchment. We also applied hydrograph separation using d18O values of water to evaluate the relative contributions of ‘‘new’’ and ‘‘old’’ water to discharge and the relation between new water and sulfate concentrations and isotopic composition.
Methods Site description Sleepers River Research Watershed (Figure 2) in northeastern Vermont, USA, was established in 1958 by the U.S. Department of Agriculture, Agricultural Research Service, and is now operated by the U.S. Geological Survey (USGS). Sleepers River was the site where the saturation-excess overland flow theory of
164
Figure 1. Sulfur cycling in a forested ecosystem and resultant changes (in italics) in S and O isotope ratios of sulfate. BSR means bacterial sulfate reduction. The isotope ratio values are given for typical precipitation in northeastern USA. Processes within the regolith are paired horizontally at the approximate depth in the system where they are most prominent. In general the process on the right immobilizes aqueous sulfate and the process on the left makes it available for aqueous transport, though in the case of dissimilatory BSR the sulfide may remain mobile, either in gas or aqueous phase. Also, this latter process often occurs near the surface in saturated areas. Activity of cosmogenic 35S (half-life 87 days) is highest in deposition, zero in weathering sources, and anywhere in between within the soil zone depending on S residence time.
Figure 2. Sleepers River Research Watershed, with enlargement of W-9 catchment, indicating sulfur isotope sampling sites.
165
166 streamflow generation was developed by Dunne and Black (1970a, b; 1971). Their work formed the basis for recent research on streamflow generation processes using chemical and isotopic approaches (Kendall et al. 1999; Shanley et al. 2002a). The 11,125-ha basin contains several gaged watersheds of various size and land cover (Shanley et al. 2002a). This study was conducted primarily within the 41-ha forested W-9 catchment and its tributary subcatchments, but samples were also taken from the 59-ha agricultural W-2, and the larger mixed land use watersheds W-3 (836 ha) and W-5 (11,125 ha). The Sleepers River watershed is underlain primarily by the Waits River Formation, a calcareous granulite interbedded with sulfidic micaceous phyllites and biotite schists (Hall 1959; Bailey et al. 2004). The bedrock is mantled with one to several meters of silty basal till derived mostly from the Waits River Formation. The calcareous lithology generates high-pH calcium bicarbonate sulfate waters throughout the Sleepers River watershed. Field sampling The study period encompassed the late winter and spring snowmelt of 2000, from late February to late April. We collected 3–12 samples from each of 10 streams draining watersheds with areas from 2 to 11,125 ha, at the same times and sites as samples collected for another study (Shanley et al. 2002b). We also sampled 3 groundwater seeps and 4 shallow wells screened within the till (1.5-m screens, maximum depth 3 m) within one day of peak snowmelt (4 to 5 April). Stream sampling included a premelt sample before the hydrograph rise, and then focused on high flows from both radiational melt and rain-on-snow events. Samples were taken from the centroid of flow in 10-L collapsible polyethylene (PE) containers for determination of 35S activity and in 1-L PE bottles for determinations of the d34S and d18O values of sulfate. A 500-ml PE bottle was filled for major anion and DOC analysis. Samples for analysis of d18O of water were collected in 16-ml glass vials with conical polyseal caps. Meltwater was collected from snowmelt lysimeters to use as the atmospheric end member in isotopic hydrograph separation (Shanley et al. 2002a). Depthintegrated snowpack samples were collected by removing an approximately 0.2 · 0.2 m column from the wall of an excavated pit with a lexan shovel into a large PE container. Four open-field snowpack samples were taken near the W-9 catchment on 23 February (just prior to a minor thaw) and two additional samples were taken on 22 March (just prior to the onset of the main melt sequence) for determination of sulfate concentrations, d34S and d18O values of sulfate, and 35S activity (22 March only). Isotopic and chemical analyses Samples for 35S activity determination were dripped through ion exchange resin columns in the field. The 35S columns and water samples for d18O analysis
167 were shipped to the USGS isotope lab in Menlo Park, CA. Sulfate was eluted from the columns and analyzed for 35S as described in Michel et al. (2000). Replicate precision averaged 0.71 mBq l1 or 0.12 mBq/mg SO4 (C.V. 33%). The d18O of water was determined with a precision of ±0.05&. Samples for anion analysis were filtered (0.45 lm) and chilled, and shipped to the USGS in Boulder, CO, where they were analyzed by ion chromatography. Samples for DOC were filtered (0.7-lm glass fiber) to amber glass vials and analyzed by ultraviolet persulfate oxidation with infrared detection at the USGS laboratory in Troy, NY. Samples for determining the stable isotope composition of sulfate were dripped through separate ion exchange columns in the field laboratory, and columns were sent to the Isotope Science Laboratory (ISL) at the University of Calgary (Alberta, Canada). In Calgary the sulfate was eluted from the column and precipitated as BaSO4 by adding a BaCl2 solution. The filtered, washed and dried BaSO4 was converted to sulfur dioxide (SO2) in an elemental analyzer and swept with a He stream into an isotope ratio mass spectrometer (VG Prism II) for isotope ratio determinations. For oxygen isotope analyses on SO42, BaSO4-oxygen was converted to CO at 1450C in a pyrolysis reactor (Finnigan TC/EA). The resultant gas was subsequently swept with a He stream into a mass spectrometer (Finnigan MAT delta plus XL) for isotope ratio determinations in continuous-flow mode (CF-IRMS). Stable isotope ratios are reported in the usual d notation in permil with respect to the international standards V-CDT for sulfur isotope measurements and V-SMOW for oxygen isotope measurements. Precision (standard deviation on replicate analyses) averaged 0.26& for d34S values and 0.53& for d18O values of sulfate. Mixing models d18O of water Snowmelt is an ideal time for isotopic separation of streamflow into event and pre-event water and the method has been successfully applied at Sleepers River (Shanley et al. 2002a). As in past studies, streamwater was separated into old and new water with a standard two-component d18O mixing model using the varying composition of meltwater from the lysimeters for the new water signal and a constant pre-melt base flow for the old water signal. d34S of sulfate We applied a two-component sulfate-S source separation to differentiate atmospheric from mineral-derived sulfate. This separation requires three conditions: (1) a constant known d34S value for atmospheric sulfate; (2) a constant known d34S value for sulfate from mineral weathering that is different from that of the atmospheric source; and (3) processes that fractionate sulfur isotopes such as dissimilatory bacterial sulfate reduction do not occur.
168 We used snowpack d34S for the atmospheric end member and stream base flow d34S for the mineral end member, giving a 2–3& difference between atmospheric and stream base flow d34S values of sulfate that was small but sufficiently distinct to satisfy condition (2). The fraction of atmospheric S in streamwater was given by: SATM ¼ ðd34 SBF d34 SSW Þ=ðd34 SBF d34 SSP Þ
ð1Þ
where subscripts ATM = atmosphere, SW = streamwater, BF = streamwater at base flow, and SP = snowpack. d18O of sulfate The d18O of sulfate was treated identically to the d34S of sulfate, assuming twocomponent mixing and using an equation analogous to equation 1. 35
S activity The specific activity of 35S (mBq per mg SO4) data were analyzed in two ways. In the first approach, we attributed all of the 35S activity in streamwater to the melting snowpack (and subsequent rainfall). This would yield a maximum contribution from the snowpack as any remaining 35S activity from aged atmospheric sulfur would be attributed to the snowpack. Mixing fractions were calculated as for d34S above, with the simplification that the 35S activity at baseflow is zero so the analagous equation reduces to: SNEW ¼ ða35 SSW =a35 SSP Þ
ð2Þ
where SNEW is the fraction of new atmospheric S in streamwater and a refers to the specific 35S activity. The second approach was to calculate the apparent age of stream sulfate assuming simple radioactive decay of 35S in precipitation: A ¼ T1=2 lnða35 SSP =a35 SSW Þ
ð3Þ
where A is mean streamwater sulfate age in days and T1/2 is the half-life of S (87 days). After the atmospheric fraction of streamwater sulfate was determined by Equation (1), its specific 35S activity was recomputed and Equation (3) was re-applied to compute the age of the atmospheric fraction of stream sulfate. 35
Results Hydrology Water Year 2000 (October 1999 through September 2000) at Sleepers River featured the typical autumn flow increase with some high flow events, followed by recession to winter base flow as precipitation was stored in the snowpack
169
Figure 3. Daily precipitation amount and continuous stream discharge hydrograph for the 2000 water year at the 41-ha W-9 catchment at Sleepers River (Figure 2).
(Figure 3). Snowmelt in March and April clearly dominated the annual hydrograph. In 2000 the snowmelt recession lasted well into June, extended by above-average rainfall in April and May. The summer of 2000 was dry, with the lowest flow of any of the nine summers on record to that point, though there were occasional moderate peak flows from thunderstorms in August and September. The winter of 2000 was relatively mild with below-average snowfall. The peak snow water equivalent (SWE) ranked 33 among 43 years of record from 1960 to 2002 (Shanley et al. 2002c). The winter was also unusual in that it lacked a major thaw, but several minor warm-ups periodically reduced the snow water equivalent and produced moderate streamflow. The main melt sequence began in late March (Figure 4). Flow progressively increased with a series of diurnal melts, culminating in two rain-on-snow events, 22 mm on 28 March and 29 mm on 4 April. These two storms generated the highest flows of the melt period (1.2 and 1.3 mm h1, respectively, in W-9, the headwater catchment), and isotope sampling focused on these and subsequent snowmelt peaks. These snowmelt peaks were less than those of most previous years, e.g., a 2.0 mm h1 peak in 1996 (Kendall et al. 1999).
Solute chemistry The sulfate concentration in the snowpack was 12 leq l1 at its maximum SWE of 199 mm on 23 February, just before a late winter thaw. On 22 March SWE was 169 mm and the sulfate concentration had decreased to 6.2 leq l1. Sulfate concentration in streamwater varied inversely with
170
Figure 4. Daily precipitation with continuous hydrograph and hydrograph separation using d18O of water during the main 2000 snowmelt period, with sulfate, DOC, and nitrate concentrations. Isotopic hydrograph separation was not possible after isotopically heavy rain fell on April 4.
Table 1. Sulfate mass balance at w-9 for water years 1992–1994 and for water year 2000; sulfate and 35S mass balance at w-9 for the 2000 snowmelt period. Units Water year 1992-1994 (Hornbeck et al. (1997)) Water year 2000 2000 snowmelt: 22 March–29 April 35 S, 2000 snowmelt, 25 March–4 April
meq meq meq mBq
2
1
m year m2 year1 m2 m2
Input
Output
Ratio
45.1 48.1 9.1 473*
106.9 131.3 41.4 157
2.37 2.73 4.55 0.33
*363 mBq m2 from snowpack; 110 mBq m2 from rainfall during this period, assuming same 35S concentration as snowpack.
discharge (r2 = 0.52, p